EmotioNL Demo: A Framework for Dutch Emotion Detection in Crisis Communication

Identifying fine-grained emotions is imperative in the context of crisis communication, e.g., in view of reputation management or monitoring the evolvement of a crisis. As crisis situations often lead to an overload of (online) information, automatic methods are crucial. Advances in natural language processing have led to sophisticated emotion detection models based on transformer models. For Dutch, a state-of-the-art emotion detection model has been created in the EmotioNL project [1]. We show how the EmotioNL model can be used in the context of crisis communication by presenting a demo hosted on Huggingface [2]. Four recent crisis situations in Belgium and the Netherlands from four distinct areas of crisis communication are included in the demo, namely the childcare benefits scandal (political crisis), the floodings from Summer 2021 (natural disaster), the COVID-19 pandemic (health crisis) and the controversies about the talent show The Voice of Holland in 2022 (organizational crisis). We show how the EmotioNL model can be used to identify emotions in tweets about those crisis situations. Furthermore, we illustrate that emotion detection can provide compelling opportunities in the field of crisis communication research: we showcase how the model can be integrated in a monitoring tool which signals significant fluctuations in emotion distributions, and demonstrate that the emotions conveyed in tweets are linked to specific topics.

[1] De Bruyne, L., De Clercq, O., & Hoste, V. (2022). Prospects for Dutch emotion detection: Insights from the new EmotioNL dataset. Computational Linguistics in the Netherlands Journal, 11, 231–255.

[2] https://huggingface.co/spaces/lunadebruyne/EmotioNL